Lessons for Preparedness and Reasons for Concern From the Early COVID-19 Epidemic in Iran

On 19 February 2020, Iran confirmed its first deaths from COVID-19 in the city of Qom. The announcement seemed to mark the beginning of the country's epidemic. Viruses do not wait for recognition. SARS-CoV-2 often causes mild illness or no obvious symptoms, allowing infected travellers to move through airports and households before a health system knows what to look for. By the time the first cases appeared in the official record, the epidemic had probably been growing for weeks.

I was the fourth-listed author in a large international collaboration that reconstructed this hidden beginning. No single dataset could reveal it. Diagnostic testing was limited, and reported cases showed only the infections that reached the surveillance system. We combined viral genomes with air-travel data, all-cause mortality and an epidemic model. Each source observed a different shadow cast by the same outbreak. Where their estimates converged, a history emerged that was difficult to see in real time.

The genomes carried the first clues. A virus copies its genetic material whenever it reproduces, and small changes sometimes enter the sequence. Viruses sampled from different patients can therefore be arranged by genetic similarity, much as related documents can be traced through shared edits. The team analysed a globally representative set of 802 genomes, including 46 from people sampled in Iran or with travel links to the country. The pattern pointed to at least five independent introductions during 2020.

Those introductions did not contribute equally. The epidemic appears to have been seeded largely by a small number of cases linked to frequent travel from China. Genetic analysis placed the ancestry of the relevant viral group well before Iran's first announcement, although the interval around that estimate was broad. Genomes evolve unevenly and early samples were sparse. The result identified a plausible window rather than the date of one decisive arrival.

Travel data offered a second clock. Infected travellers detected after leaving Iran carried information about how common infection must have been among the population from which they departed. The study used 36 exported cases to estimate prevalence and placed the epidemic's start around 25 December 2019, with a wide 95 per cent interval. Early cases may have doubled roughly every four days. Such growth can remain inconspicuous at first, then become overwhelming within several doublings.

The calculation depended on assumptions about who travelled and how likely an infection was to be detected abroad. Air passengers may differ from the wider population in age, wealth or contact patterns. Destination countries also varied in surveillance. The exported cases still provided valuable evidence during a period when domestic testing revealed only a small fraction of infections. Mobility had carried the virus and, paradoxically, helped measure its hidden spread.

Exponential growth explains why a short delay mattered so much. If infections double every four days, one chain becomes roughly thirty times larger in less than three weeks when nothing slows it. Early numbers can remain small enough to ignore while the next generations are already forming. By the time hospitals detect the rise, transmission reflects decisions made many days earlier. Surveillance buys value by shifting recognition closer to the start of that sequence.

Mortality showed the outbreak from another direction. The researchers compared deaths from all causes with the number expected from earlier years. By the end of summer 2020, they estimated about 21,900 excess deaths, with an interval from 16,700 to 27,200. Some excess deaths may have arisen indirectly when ordinary care was disrupted, while confirmed COVID-19 deaths missed people who were never tested. The gap between the measures nevertheless indicated substantial under-reporting and a burden far larger than the official case count implied.

Under-reporting in this setting described the limits of observation more than one single act of concealment. Mild infections never reached a clinic, tests were scarce and death registration took time. Diagnostic criteria changed as knowledge improved. Political pressures could affect reporting too, but the analysis did not need to assign every missing case a motive. It showed that several independent measures were inconsistent with the smaller epidemic visible in the official series.

These streams entered an SEIR model, which represents people moving through stages of susceptibility, exposure, infectiousness and recovery. The compartments simplify millions of individual lives so that researchers can test whether an epidemic curve is consistent with the available evidence. In this case, the reconstruction suggested that about 15 million people might have recovered by 31 August 2020, again with considerable uncertainty. The value lay in estimating scale and direction when direct observation was incomplete.

The model also examined interventions that reduce contact without using medicines or vaccines. Restrictions on gatherings and travel, isolation and changes in daily behaviour can lower the number of new infections generated by each case. Iran's measures slowed transmission. When restrictions eased while many people remained susceptible, the model anticipated another peak. Resurgence was a consequence of the remaining conditions, rather than proof that the earlier measures had achieved nothing.

Health-system capacity formed the human reason for slowing the curve. Even when the same number of people eventually becomes infected, spreading severe cases over a longer period can keep demand closer to the available beds and staff. It also creates time to improve clinical practice and secure supplies. A modelled epidemic curve becomes a clinical reality through the number of patients arriving together. The height of the peak determines whether each can receive the care the system knows how to provide.

Border screening had clear limits. Temperature checks can miss an infected person before fever begins, and many people never develop one. Travel restrictions may delay introduction, which can be valuable when the time is used to expand testing or prepare hospitals. Once community transmission is established, concentrating surveillance at the border leaves the larger problem inside the country. The study's evidence from several introductions supported layered preparedness instead of confidence in a single checkpoint.

Genomic surveillance later became a familiar part of pandemic reporting, but its usefulness depends on representative sampling. Sequences concentrated in one city or among international travellers may miss transmission elsewhere. Laboratories need a route for linking genetic findings with dates and broad epidemiological information while protecting privacy. A genome becomes most informative when it can be interpreted beside the circumstances of the infection it represents.

Mortality surveillance deserves the same long-term investment. All-cause deaths can reveal a crisis even when the diagnosis is disputed or testing fails. Timely registration matters, since figures released many months later cannot guide an unfolding response. Analysts also need stable historical records to establish a credible baseline. The infrastructure that seems administrative during ordinary years becomes a form of emergency detection when a new pathogen arrives.

The Iranian reconstruction illustrates why uncertainty should prompt action rather than paralysis. Early estimates covered wide ranges, yet several sources agreed that transmission had started earlier and grown larger than recognised. Waiting for a precise count would have allowed exponential growth to continue. Preparedness uses the best current estimate, states what could change it and chooses measures robust to several plausible realities.

Communicating a range is part of that discipline. A central estimate can be repeated as though it were an observed count, stripping away the uncertainty that gives it meaning. Decision-makers need to understand whether every plausible value leads toward the same action. In Iran, the exact starting date remained uncertain while the evidence for substantial unseen transmission was strong. The response could proceed on that shared implication without pretending the model knew the unknowable.

International cooperation was part of that preparedness because infected travellers linked epidemics across borders. Rapid sharing of sequences and anonymised travel information helped researchers see a pattern no national dataset contained alone. That exchange has to remain reciprocal. Countries that report an emerging threat should gain technical support rather than stigma or automatic punishment, which can encourage delay the next time an unusual cluster appears.

Iran's first recognised cases were therefore a late chapter in the opening story. Viral lineages preserved hints of earlier movement. Departing passengers revealed infections that domestic testing had missed, and mortality records showed the weight of the unseen epidemic. Brought together, the evidence replaced a sudden beginning with weeks of quiet growth. The lesson for the next outbreak is simple in principle and demanding in practice: a country needs several ways of seeing before the threat has a name.